0

Notes on the Outlook.com AS9031519 bounces (Nov 5-9): the real start date, why M365 NDRs blamed SPF, and how to find what bounced

preview

I run a small DMARC SaaS, so the “550 5.7.1 Message rejected: Message lacks SPF/DKIM/DMARC Auth Records (AS9031519)” bounces took over my week. Microsoft has since called it a fault and rolled it back. Here is what I pieced together from Microsoft’s bulletin, the Microsoft Q&A threads and a few operators’ logs, plus the part that still needs doing. 𝗪𝗵𝗮𝘁 𝗠𝗶𝗰𝗿𝗼𝘀𝗼𝗳𝘁 𝘀𝗮𝗶𝗱 Service Health incident EX1545193 (tenant admins only): “A recent update to the spam enforcement rules caused email messages to be incorrectly marked and blocked, which led to impact.” Window: Nov 6 18:00 UTC to Nov 9 11:02 UTC. The bulletin never mentions DMARC, and it expands SPF as “Service Provider Foundation”. 𝗜𝘁 𝘀𝘁𝗮𝗿𝘁𝗲𝗱 𝗯𝗲𝗳𝗼𝗿𝗲 𝗠𝗶𝗰𝗿𝗼𝘀𝗼𝗳𝘁’𝘀 𝘄𝗶𝗻𝗱𝗼𝘄

  • Nov 5: rejection appears in operator logs, “not once before that” (Mythic Beasts, MailRoute)
  • Nov 6 06:19 UTC: earliest Microsoft Q&A report I found (Microsoft Q&A)
  • Nov 6 09:39:53 UTC: bounce stamped by Microsoft’s own server, AS9031519 included (Lazy Admin)
  • Nov 6 18:00 UTC: Microsoft’s stated start (EX1545193)
  • Nov 9 11:02 UTC: Microsoft’s stated end (EX1545193) A log search that starts at Microsoft’s start misses Nov 5 and the morning of Nov 6. 𝗪𝗵𝘆 𝗠𝟯𝟲𝟱 𝗡𝗗𝗥𝘀 𝘀𝗮𝗶𝗱 “𝗦𝗲𝗻𝗱𝗲𝗿 𝗣𝗼𝗹𝗶𝗰𝘆 𝗙𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 𝘃𝗶𝗼𝗹𝗮𝘁𝗶𝗼𝗻” M365 senders got 550 5.7.23 The message was rejected because of Sender Policy Framework violation -> 550 5.7.1 Message rejected: …. 5.7.23 is in Exchange Online’s own NDR table; Postfix and Zoho senders only ever saw the raw 5.7.1. My reading: the SPF wording came from Exchange Online’s NDR, not from a failed SPF check. Microsoft hasn’t said it in those words. 𝗪𝗵𝘆 𝘀𝗼𝗺𝗲 𝗵𝗲𝗮𝗱𝗲𝗿𝘀 𝘀𝗮𝗶𝗱 𝗱𝗺𝗮𝗿𝗰=𝗽𝗮𝘀𝘀 𝘄𝗶𝘁𝗵 𝗻𝗼 𝗿𝗲𝗰𝗼𝗿𝗱 Microsoft writes dmarc=bestguesspass when no DMARC record exists (“If the domain had a DMARC TXT record, the DMARC check for the message would pass.”). That is most likely what the “but my headers say pass” posts were reading. Check DNS instead: dig +short TXT _dmarc.example.com. 𝗜𝘁 𝘄𝗮𝘀𝗻’𝘁 𝗗𝗠𝗔𝗥𝗖 𝗲𝗻𝗳𝗼𝗿𝗰𝗲𝗺𝗲𝗻𝘁 RFC 9989 §4.10.1: with no policy record, receivers “MUST NOT apply the DMARC mechanism”. This was Outlook.com’s own filtering. The Postmaster policies still only strongly recommend DMARC below 5,000 emails a day. 𝗪𝗵𝗮𝘁 𝘀𝘁𝗶𝗹𝗹 𝗻𝗲𝗲𝗱𝘀 𝗱𝗼𝗶𝗻𝗴- Find the bounces from Nov 5 00:00 UTC to Nov 9 11:02 UTC. Match the text, not the code: 5.7.1 is generic, and day one had no AS9031519 suffix.
  • Postfix (current and rotated logs): zgrep -h ‘Message lacks SPF/DKIM/DMARC Auth Records’ /var/log/mail.log* | grep -o ‘to=<[^>]*>’ | sort | uniq -c | sort -rn
  • M365: EAC → Mail flow → Message trace, recipients *@outlook.com, *@hotmail.com, *@live.com, *@msn.com (one wildcard per value), delivery status Failed. Trace reaches back 90 days.
  • Resend what matters. It was a 550, so nothing was retried.
  • Keep the emergency p=none and add rua= so you get aggregate reports.

𝗪𝗵𝗮𝘁 𝗜 𝗱𝗼𝗻’𝘁 𝗸𝗻𝗼𝘄 Some people said they bounced with SPF, DKIM and DMARC all in place. Unverified, and several Q&A cases turned out to have no record after all. Nobody publicly tested a domain with a DMARC record but no aligned SPF or DKIM. If that’s you, keep the bounce and the headers before changing anything. Happy to look at a redacted bounce if yours doesn’t fit any of this. Full write-up with all four bounce forms, the sources and the timeline: dmarcguard.io/…/outlook-as9031519-dmarc-bounce/

I run a small DMARC SaaS, so the “550 5.7.1 Message rejected: Message lacks SPF/DKIM/DMARC Auth Records (AS9031519)” bounces took over my week. Microsoft has since called it a fault and rolled it back. Here is what I pieced together from Microsoft’s bulletin, the Microsoft Q&A threads and a few operators’ logs, plus the part that still needs doing. 𝗪𝗵𝗮𝘁 𝗠𝗶𝗰𝗿𝗼𝘀𝗼𝗳𝘁 𝘀𝗮𝗶𝗱 Service Health incident EX1545193 (tenant admins only): “A recent update to the spam enforcement rules caused email messages to be incorrectly marked and blocked, which led to impact.” Window: Nov 6 18:00 UTC to Nov 9 11:02 UTC. The bulletin never mentions DMARC, and it expands SPF as “Service Provider Foundation”. 𝗜𝘁 𝘀𝘁𝗮𝗿𝘁𝗲𝗱 𝗯𝗲𝗳𝗼𝗿𝗲 𝗠𝗶𝗰𝗿𝗼𝘀𝗼𝗳𝘁’𝘀 𝘄𝗶𝗻𝗱𝗼𝘄

  • Nov 5: rejection appears in operator logs, “not once before that” (Mythic Beasts, MailRoute)
  • Nov 6 06:19 UTC: earliest Microsoft Q&A report I found (Microsoft Q&A)
  • Nov 6 09:39:53 UTC: bounce stamped by Microsoft’s own server, AS9031519 included (Lazy Admin)
  • Nov 6 18:00 UTC: Microsoft’s stated start (EX1545193)
  • Nov 9 11:02 UTC: Microsoft’s stated end (EX1545193) A log search that starts at Microsoft’s start misses Nov 5 and the morning of Nov 6. 𝗪𝗵𝘆 𝗠𝟯𝟲𝟱 𝗡𝗗𝗥𝘀 𝘀𝗮𝗶𝗱 “𝗦𝗲𝗻𝗱𝗲𝗿 𝗣𝗼𝗹𝗶𝗰𝘆 𝗙𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 𝘃𝗶𝗼𝗹𝗮𝘁𝗶𝗼𝗻” M365 senders got 550 5.7.23 The message was rejected because of Sender Policy Framework violation -> 550 5.7.1 Message rejected: …. 5.7.23 is in Exchange Online’s own NDR table; Postfix and Zoho senders only ever saw the raw 5.7.1. My reading: the SPF wording came from Exchange Online’s NDR, not from a failed SPF check. Microsoft hasn’t said it in those words. 𝗪𝗵𝘆 𝘀𝗼𝗺𝗲 𝗵𝗲𝗮𝗱𝗲𝗿𝘀 𝘀𝗮𝗶𝗱 𝗱𝗺𝗮𝗿𝗰=𝗽𝗮𝘀𝘀 𝘄𝗶𝘁𝗵 𝗻𝗼 𝗿𝗲𝗰𝗼𝗿𝗱 Microsoft writes dmarc=bestguesspass when no DMARC record exists (“If the domain had a DMARC TXT record, the DMARC check for the message would pass.”). That is most likely what the “but my headers say pass” posts were reading. Check DNS instead: dig +short TXT _dmarc.example.com. 𝗜𝘁 𝘄𝗮𝘀𝗻’𝘁 𝗗𝗠𝗔𝗥𝗖 𝗲𝗻𝗳𝗼𝗿𝗰𝗲𝗺𝗲𝗻𝘁 RFC 9989 §4.10.1: with no policy record, receivers “MUST NOT apply the DMARC mechanism”. This was Outlook.com’s own filtering. The Postmaster policies still only strongly recommend DMARC below 5,000 emails a day. 𝗪𝗵𝗮𝘁 𝘀𝘁𝗶𝗹𝗹 𝗻𝗲𝗲𝗱𝘀 𝗱𝗼𝗶𝗻𝗴- Find the bounces from Nov 5 00:00 UTC to Nov 9 11:02 UTC. Match the text, not the code: 5.7.1 is generic, and day one had no AS9031519 suffix.
  • Postfix (current and rotated logs): zgrep -h ‘Message lacks SPF/DKIM/DMARC Auth Records’ /var/log/mail.log* | grep -o ‘to=<[^>]*>’ | sort | uniq -c | sort -rn
  • M365: EAC → Mail flow → Message trace, recipients *@outlook.com, *@hotmail.com, *@live.com, *@msn.com (one wildcard per value), delivery status Failed. Trace reaches back 90 days.
  • Resend what matters. It was a 550, so nothing was retried.
  • Keep the emergency p=none and add rua= so you get aggregate reports.

𝗪𝗵𝗮𝘁 𝗜 𝗱𝗼𝗻’𝘁 𝗸𝗻𝗼𝘄 Some people said they bounced with SPF, DKIM and DMARC all in place. Unverified, and several Q&A cases turned out to have no record after all. Nobody publicly tested a domain with a DMARC record but no aligned SPF or DKIM. If that’s you, keep the bounce and the headers before changing anything. Happy to look at a redacted bounce if yours doesn’t fit any of this. Full write-up with all four bounce forms, the sources and the timeline: dmarcguard.io/…/outlook-as9031519-dmarc-bounce/